Online Program Home
My Program

Abstract Details

Activity Number: 254 - Contributed Poster Presentations: Section on Statistical Learning and Data Science
Type: Contributed
Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #328980
Title: Global Sensitivity Analysis from Given Data : Elementary Effect Approach
Author(s): Jong hyun Kim* and Dae il Jang and Kyung joon Cha
Companies: Hanyang University and Hanyang University and Hanyang University
Keywords: Global sensitivity analysis; Elementary Effect method; simplex-based design; variable screening; metamodel

While global sensitivity analysis (GSA) is basically a model-based study which reveals the relation between uncertainty of model output variable and input variables, in some cases GSA should be performed in a "data-given situation", which means the analyst cannot choose the sample points. In this situation, the analyst may construct a metamodel (or emulator) using the given data, and then conduct ordinary model-based GSA methods. However, these approaches can have significantly sensitive effects on the analysis depending on the quality of the model, and the results may even vary entirely depending on which modelling method to be chosen. In this study, we introduce a direct GSA screening method for given data, without building a metamodel. The main idea of this method is to form a data-simplex which can directly applied to Elementary-Effect (EE) method, using the concept of EE proposed by Morris in 1991 and the simplex-based sampling proposed by Pujol in 2009. Another advantage of this method is that GSA is performed in the pre-modelling stage, which can provide important information for data driven modelling, such as variable screening or model selection.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program